Serum concentration of carbohydrate 19-9 antigen (CA 19-9), a sensitive marker of pancreatic cancer and cholangiocarcinoma, increases in a variety of liver diseases due to higher production and release by bile duct cells. Biliary epithelial cells are primarily affected in liver disease associated with cystic fibrosis (CF), which develops in up to 30% of CF patients. Our aim was to evaluate the usefulness of serum CA 19-9 concentration as a test of liver involvement in CF. Serum concentration of CA 19-9, liver enzymes, bile acids, and total amylase was determined in 107 CF patients (49 with and 58 without liver disease) and 56 healthy subjects. Serum CA 19-9 concentration was significantly higher in CF patients (67 U/ml, 95% CI 53.5-80.5 U/ml) than in controls (11.8 U/ml, 95% CI 2.5-44 U/ml; p < 0.001) and in CF patients with liver disease (92.3 U/ml, 95% CI 75-109.5 U/ml) compared to CF patients without liver disease (46.6 U/ml, 95% CI 27.8-65.4 U/ml; p < 0.001). In CF patients, stepwise logistic regression analysis identified alanine aminotransferase, gamma-glutamyltranspeptidase, amylase, and CA 19-9 as the most useful predictors of liver disease (p < 0.0001). Receiver-operating characteristic (ROC) curve analysis revealed gamma-glutamyltranspeptidase and CA 19-9 as the best tests for identification of liver disease in CF patients; at a CA 19-9 cut-off arbitrarily fixed at 73 U/ml, positive and negative predictive value was 70% and 78%, respectively (sensitivity 57%, specificity 81%). To increase sensitivity to 94%, the cut-off had to be fixed at 31 U/ml, which corresponds to a specificity of only 36.2% (predictive value 33%). Our study indicates that measurement of the serum CA 19-9 concentration alone cannot be proposed as a reliable test of early hepatic involvement in CF.